Depth estimation of the light field image based on convolutional neural network
Pei Qiao, Deyuan Li, Xuewen Yan, Hua Li, Mengqing Niu · Eighth Symposium on Novel Photoelectronic Detection Technology and Applications · 2022
Light field imaging uses a single camera to capture an array of images from different angles. It can obtain the spatial information and angle information of the light scene by placing a micro-lens array between the lens and the image sensor to, which is widely used in computer vision, industrial inspection, medical imaging. The image depth estimation algorithm affects the spatial resolution of reconstructed images. A convolution neural network algorithm based on Epipolar Image (EPI) was developed and optimized by adjusting the parameters. i.e. learning rate and activation function. The depth information of light field image was predicted with an accuracy of 67.5%. The convolution neural network algorithm effectively improved the efficiency of the depth estimation and realized the 3D image reconstruction.